Will Wright on Prototypes
Edited from the transcript for readability:
Just to illustrate the differences between the two modeling techniques, in traditional math if we had let's say a square mile of land and we planted some crops on it standard math would probably go in here, measure the size of each of these things that ellipsoid would be kind of problematic and there'd be a kind of a long equation that you'd run all the numbers through to calculate what area of this plot of land was covered by fields.
Using modeling techniques we probably do something much stupider, we start throwing darts at it randomly and after we threw a certain number of darts we would then measure how many of the darts landed in a field and how many landed outside of a field and that ratio would tell us what percentage of the area was occupied by crops.
As we throw more darts we get a more accurate answer and if you can see this in some sense on the surface is a kind of a stupid approach but using the power of a computer this can actually give you very good results.
This is called by the way the monte carlo technique for kind of obvious reasons. This is a stochastic method which in simulation stochastic means there's some amount of randomness involved.
The same run of the model won't always give the same exact result, so basically as designers we have these dynamic spaces. There's this vast landscape of dynamic spaces that we can put into our games and we cut off little chunks of that space with rulesets, then we build our games around.
So as a designer one of the things I like to do is explore this space and figure out which dynamics are out there that would make for interesting game play.
Part of this talk I'm going to be using prototypes to show you examples of how we map this space.
The way I view prototypes basically is the same as paratroopers, we just drop these little paratroopers where we think might be interesting spots on the landscape, when they land we can start playing with the prototype and seeing how interesting was that space actually and then start iterating the prototype uphill in regions and so the prototypes in some sense are kind of hill climbing things in local regions of this dynamic space.